Page 153 - Proceedings of the 2018 ITU Kaleidoscope

Machine learning for a 5G future

ethical and moral understanding, this vision may never erased or hacked. If it get disturbed than machine should be
become true for a designer and hence for her machine too. put in shut down mode [23].
However, this is what one can aim for the machines to
move closer to behaving as sentient beings, and yet ensure At the organizational level another aspect that requires
that they not digress or regress. attention is that most AI solutions suffer from ―Black Box
Phenomenon‖[42] and an important step to tackle this
6. IMPLEMENTATION STRATEGIES would be to ―open the black box‖. This can be achieved by
the documentary auditing of algorithms and databases
Successful implementation of the proposed framework posteriori[44].
entails involvement of all the stakeholders from the highest
national policy making organs of the government to the 7. SUGGESTED WAY FORWARD
operational level implementation at the policy. Globally,
governments across countries are seized of the need to Since ‗ethics‘ must become the cornerstone of AI/ ML in
have national level policies as well as bodies looking into the global digital economy, there must be a united global
the aspects of ethics in ML that can advise the governments efforts in this direction. First and foremost, there is a need
on these complex issues. At the national level, they may be for an ethical framework, which can serve as a standard for
designated as Centre of Research Excellence for AI imbuing the ‗Ethical Framework‘ in all intelligent
(COREs)[42] or Centre for Data Ethics and Innovation [43] machines in the world. For this, some international
and their mandate would be to work with experts drawn guidelines may be formulated by a consortium of global
from different fields to develop an ethical framework. The leaders. This global consortium also be empowered to
ethical framework proposed in section 5 is at that higher monitor an insistent inculcation (with minor contextual
level of thinking. Its implementation will however, require modifications) of these guidelines by all the related
that this framework is publicized; policy changes are organisations and governments involved in research and
suggested covering private as well as public organisations; development of AI. Taking a cue from these global
promotion of standards around the use of data; encourage guidelines, appropriate interventions are required at the
research in emerging areas; and make guidelines focusing national level too. For instance, there must be appropriate
on incorporating the framework at the design stage, regulatory framework, standards and templates in each
actively encourage transparency and auditability of the country to enable integration of the proposed conceptual
systems. framework /design approach in the conventional approach
of ML programming. There is also a need of a national
The next step in the implementation hierarchy would be at authority as brought out in section 6 that periodically
the organizational level. This would require changes in the checks ethical indices of the various ML innovations in the
hiring, training/capacity building, promotion and feedback country. Countries also need to establish a formal as well as
policies of the organisations. The organisations would an informal network of distinguished think tanks,
consciously need to focus on hiring persons who fit in the academics, ethic advocates, practitioners and industry
quadrant 2 &3 of the framework depending upon their professionals, who could also have knowledge linkages
requirement and also keeping in view the overall national with a similar consortium at global level. Such a cohesive
policies governing this work. Incorporating ethical aspects global approach would help to steer the world towards a
at the design stage would be required and there may be a more ethical model of progress that is likely to emanate
possibility of an ‗ethics engineer‘ to be part of the team from ever-mushrooming AI/ ML based innovations.
working on ML and AI solutions. Mere programming or
algorithm writing will not address the underlying issues as 8. CONCLUSIONS
ethics are a manifestation of intent as well as action. The
‗ethics engineer‘s‘ job would be to fulfill the ethical With the threat of an Artificial Intelligence arms race
requirements of the system from the time when the looming large on the future of humankind on earth, it is an
designers feed mathematical algorithms to a careful urgent requirement to speed up the process of introduction
monitoring of how the machine uses its algorithms to fulfill of ethical intelligence algorithms in machine learning.
the ethical objectives fed into it. Some of the broad Cyber Security and privacy concerns are only going to
principles that can be considered are: further get complicated with increasing autonomy in
machine learning. For the longevity of natural life on earth
Machine should be obedient to owner, designer and it is imperative we design algorithm for machines that
authority. Existing program cannot be erased without the become aware of what harmony, empathy, conservation of
authority and consent of system designer. There is need to resources is. Humans need to secure not just their future but
provide degree of autonomy to machine. So that machine also life on earth in general.
cannot erase some basic program however new cognitive
learning program can be erased after some checks. Some Nevertheless, addition of spiritual and emotional quotient
basic program should be always locked and cannot be in the AI might also lead the machines to begin questioning
existential questions that humans may already be facing

– 137 –